This chapter examines naturalistic accounts of mental disorder: accounts that define disorder as biological dysfunction. There are three such accounts: an eliminativist account ; a forward-looking or goal-contribution account and a backward-looking or evolutionary account. I argue first, and contra Szasz, that biological functions can be attributed at a mental level. But our mental architecture might simultaneously support many different ways of attributing function claims, which might undermine a strong naturalism about mental disorder. Second, I argue that Boorse's forward-looking account (...) of disorder as dysfunction is not value-free. Third, I argue that Wakefield's backward-looking account does not accurately map onto our disorder judgments or medical purposes. I conclude that whilst Boorse's forward-looking account of disorder as dysfunction tracks medical usage better than Wakefield's evolutionary account, a truly value-free account of mental disorder cannot be sustained. (shrink)